Polymarket started settling its short-dated Bitcoin markets on an averaged price on August 7, closing the loophole a Stanford University and Singapore Management University study said let traders pull $8.2 million out of the exchange's five-minute Bitcoin contracts by nudging Binance's spot price in the final ten seconds before each one closed.
The contracts ask a simple question: will Bitcoin be above or below a set line when a five-minute clock runs out. Until August 7, the answer came from one Chainlink price reading taken at the instant the window closed. Binance's own mid-price already sits within about 2.5 basis points of that reading, so a large enough order in the closing seconds could tip the outcome, then let the price snap back once the contract had settled.
How big the pattern was
The study, by researchers Shihao Yu, David Dai and Ruizhe Jia, went through roughly 16,000 of the five-minute contracts over a two-month stretch and flagged 1,613 settlement windows as likely manipulated. Directional buying or selling on Binance in the final ten seconds of those windows averaged $1.7 million, against $68,000 in an ordinary cycle. The pattern traced back to 821 accounts, about 0.3% of the roughly 243,000 that traded the contracts, who together made $8.2 million; 93% of that, about $7.6 million, came out of retail traders on the other side of the trade.
Warnings had circulated for months before the study put a number on the scale. In May, the pseudonymous trader Variance Lover posted examples of the pattern online, and Josh Stevens, an engineer at Polymarket, replied that the team was looking into it. The academic paper that followed in July confirmed what traders had only been able to describe.
The fix: average the price, don't freeze it
The same study found almost none of the pattern in Polymarket's 15-minute Bitcoin contracts, since a price push that is cheap to hold for a few seconds gets expensive to hold for a few minutes. Polymarket's fix follows that logic directly. Instead of reading Bitcoin's price at one instant, the exchange now averages it over a window: 30 seconds for the five-minute markets, 60 seconds for the 15-minute and four-hour ones, a method called a time-weighted average price, or TWAP. The feeds run on Chainlink Data Streams, which went live for this purpose on July 31, and Polymarket is putting $1 million into liquidity rewards on the affected markets through the end of August to keep them attractive to trade while the new pricing settles in.
The fix covers the specific mechanism the study tested, not every short window on the platform. Polymarket's five-minute and 15-minute Bitcoin markets and its four-hour crypto contracts now settle on an average; any other market that still resolves on a single price at a single second is exactly as fast to move as it was on August 6.
